Time Series Forecasting:
Machine Learning and Deep Learning with R and Python
- Exam Guidelines -

Marco Zanotti

2021-2022

Content

This is challenge is all about time series forecasting & reporting.

The purpose of this challenge is to verify your ability to accurately and timely forecast many business time series of different frequencies.

You are free to use the tool you prefer to estimate models and produce forecasts.

Requirements

You are required to produce a notebook to present your whole project, from methodologies used to the results obtained, carefully explaining your approaches.

In particular, you have to report: - list of forecasting methods used
- accuracy results on test set for each time series and each method using RMSSE, MASE and sMAPE
- best accuracy results on test set for each time series using RMSSE, MASE and sMAPE
- average accuracy result on test set (Average RMSSE, MASE and sMAPE)
- total computation time required to make the computations with system information
- total time spent on developing the project

Data

The data comes from the M4-Competition, the fourth of the Makridakis Competitions, a series of open competitions to evaluate and compare the accuracy of different time series forecasting methods.

The “M” competitions have had an enormous influence on the field of forecasting.
“They focused attention on what models produced good forecasts, rather than on the mathematical properties of those models.” Professor Rob J. Hyndman

The dataset of the M-Competitions are publicly available and downloadable from the International Institute of Forecasters. In particular, the exam data is taken sampling the 120 different time series from the M4-Competition at various periodicities.

You are encouraged to read about the main results of this competition:
- The M4 Competition: 100,000 time series and 61 forecasting methods
- The M4 Competition: Results, findings, conclusion and way forward

Exam dataset contains 120 time series from the M4-competitions:
- 20 hourly time series
- 20 daily time series
- 20 weekly time series
- 20 monthly time series
- 20 quarterly time series
- 20 yearly time series